That's a really interesting approach. The trick would be knowing the indicators of fabricated data. Or perhaps just knowing when the data isn't good enough to pass the next clinical trial.